ag-ui-knowledge
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a knowledge base for the AG-UI protocol and does not contain any executable scripts or malicious instructions. It provides guidance on best practices, such as using middleware for tool filtering and encrypted reasoning for privacy compliance.
- [EXTERNAL_DOWNLOADS]: The documentation includes standard instructions for installing official SDKs from NPM (
@ag-ui/client) and PyPI (ag-ui-protocol). These references are informational and point to legitimate, well-known package registries. - [INDIRECT_PROMPT_INJECTION]: The skill describes a protocol that processes user-supplied messages and tool outputs. It includes documentation on boundary markers (message roles) and security controls like
FilterToolCallsMiddlewareto mitigate risks associated with untrusted data ingestion. Mandatory Evidence Chain: 1. Ingestion points:RunAgentInput.messages(documented inconcepts/messages.md). 2. Boundary markers: The protocol uses typed events and roles to delineate content. 3. Capability inventory: Tool calls and state mutation are core documented capabilities. 4. Sanitization: The documentation explicitly referencesFilterToolCallsMiddlewarefor validating and filtering tool execution.
Audit Metadata